Characterizing the impact of interference on medium access in multi-hop wireless networks
Interference plays a complex and often defining role on the overall performance of wireless networks, especially in multi-hop scenarios. Understanding this role is critical for estimating the performance of these networks, and in turn for developing effective protocols for them. Depending on the relative location of interfering sources and destinations, several modes of interference exhibiting different behavior, occur. In this work, we start with categorizing these interference modes between two concurrent transmissions (the basic unit of interference in a network) in order to understand their behavior. We term these modes of interference as interference interactions. We identify 5 types of interactions. We develop closed form geometric models to determine occurrence frequencies for each category and verify these models against Monte Carlo simulation based results.
Our next step is to extend this work by using a more realistic model for packet reception. We use Signal to Interference and Noise (SINR) model and capture effects for packet reception which allows for a more accurate modelling of packet reception. Based on this model, which makes our analysis more complicated, our interaction categories increase to 11. We develop geometric models to predict the frequency of occurrence of these interactions. These categories allow us, for the first time, to quantify exposed node problem in a general wireless network.
We show that MAC level interactions play a primary role in determining the behavior of chains. We evaluate the types of chains that occur based on the MAC interactions between different links using realistic propagation and packet forwarding models. We discover that the presence of destructive interactions, due to different forms of hidden terminals, does not impact the throughput of an isolated chain significantly. However, due to the increased number of retransmissions required, the amount of bandwidth consumed is significantly higher in chains exhibiting destructive interactions, substantially influencing the overall network performance.
We extend our chain analysis to study how TCP connections, which involve bidirectional flows, behave over wireless chains. First, we break down and examine the types of chains that occur most frequently in TCP configurations and classify them by the nature of the MAC level interactions that arise in each. We then show that unlike uni-directional data communication that we studied earlier, the throughput of TCP over a wireless chain is greatly affected by the type of interactions within the chain. Depending on these interactions, we observe a throughput difference of up to 25 percent. Finally, we show the implications of the MAC level interactions on network performance: specifically, route instability and number of retransmissions.
We identify the cross chain interactions for the most common categories of chains. We then study the probability of occurrence of cross chain MAC interactions and their effects on the performance of chains. We also show that transmission patterns of a chain depend on the interference interactions between its links; an effect that is necessary for accurately estimating chain performance. We observe that destructive interactions arise more frequently among two interfering chains than they do within a single chain. Moreover, chains that have destructive interactions due to self-interference are more prone to having cross-chain destructive interactions. Thus, both intra-chain as well as cross-chain interactions, ultimately provide significant insight into how chains interact.
Finally we identify several applications at different layers of the network protocol stack that can benefit from the knowledge of these interactions. We show that interference interactions can be used as the guide to tweak network parameters like transmit power and carrier sense threshold. We develop an algorithm that allows us to set these parameters in order to maximize channel reuse and at the same time minimize destructive interference between links. We call this application interaction engineering as we aim to modify destructive interactions to more stable ones. (Abstract shortened by UMI.)